(Bloomberg) -- For months, researchers have believed that the very first known U.S. case of Covid-19 in mid-January may have seeded an initial large outbreak in the Seattle area, spreading undetected for almost six weeks.

However, a new analysis from researchers at the University of Arizona and elsewhere, based on computer simulations of how the virus evolves, indicates that this is most likely incorrect. Instead, the Seattle-area outbreak likely started weeks later with someone who arrived from Asia around Feb. 13, during a period in which thousands of Americans were returning from China as the outbreak there expanded.

The analysis, which was posted online and has not been peer reviewed, also suggests that an outbreak in Italy did not stem from an early German coronavirus cluster at an auto-parts maker, as some researchers have suggested. Instead, it also likely started later through an independent introduction from China.

The new study provides reason for optimism: It suggests that if Covid-19 cases can be brought down to very low numbers, it’s possible to use techniques such as contact tracing to keep an outbreak under control.

“In a lot of ways, it is good news,” said Michael Worobey, an evolutionary biologist at the University of Arizona, who led the study in collaboration with researchers at University of California San Diego, UCLA, KU Leuven in Belgium, and elsewhere. “It shows that some of the tools that we have to fight this virus actually succeeded in some of these early cases.”

The shifting conclusions are also another example of how scientists adjust their theories as more data rolls in. During the pandemic, this normal but messy process is increasingly occurring in full public view.

Having more data “helps us refine our hypotheses,” said Joel Wertheim, a molecular epidemiologist at the University of California San Diego and a co-author of the analysis. “The world is witnessing science being done in real time.”

The concept that the the first U.S. patient directly led to the Washington outbreak was popularized by Trevor Bedford, a computational biologist at the Fred Hutchinson Cancer Research Center, who originally posted his claim on Twitter on Feb. 29.

In a stream of tweets Monday, Bedford said he had changed his mind, saying that he now did not believe the first patient, dubbed “WA1,” was the beginning of the state’s outbreak. He said he still thought the virus began circulating in Washington relatively early, sometime between Jan. 18 and Feb. 9.

Bedford’s original argument was based on the idea that the first Washington patient and those in the outbreak that followed had viruses that were genetically similar, but contained a key genetic variant that was uncommon in patients in China.

The claim was compelling to Worobey as well, he said in an interview, until he started to look at the data more carefully several weeks ago. By this point, much more genomic information was available, and Worobey said the pattern of mutations didn’t make sense. Almost all the later patients in Washington had two specific mutations that weren’t present in the first Washington patient. But none of the later patients had just one of the mutations, nor were any of the viruses in the later patients identical to the original patient, as would have been expected.

Worobey said he brought in other researchers to simulate how often the mutation pattern would have occurred if the first patient really did seed the later outbreak. The researchers performed 1,000 simulations but weren’t able to reproduce the actual pattern of mutation that was seen in the Seattle outbreak. Other simulations also indicated that the Italian epidemic also was highly unlikely to have stemmed from the original German cases.

“Our results refute prior findings erroneously linking cases in January 2020 with outbreaks that occurred weeks later,” the researchers wrote in a preprint posted on the server bioRxiv. “Instead, rapid interventions successfully prevented onward transmission of those early cases in Germany and Washington State.”

©2020 Bloomberg L.P.